Class Wagging

JSAT/src/jsat/classifiers/boosting/Wagging.java:38–369

Wagging is a meta-classifier that is related to Bagging. Instead training on re-sampled data sets, it trains on randomly re-weighted data sets. The weight of each point is selected at random from a specified distribution, and set to zero if negative. See: <a href="http://www.springe

 * @author Edward Raff
 */
public class Wagging implements Classifier, Regressor, Parameterized
{

	private static final long serialVersionUID = 4999034730848794619L;
	private ContinuousDistribution dist;
    private int iterations;
    private Classifier weakL;
    private Regressor weakR;
    
    private CategoricalData predicting;
    
    private Classifier[] hypotsL;
    private Regressor[] hypotsR;

    /**
     * Creates a new Wagging classifier
     * @param dist the distribution to select weights from
     * @param weakL the weak learner to use
     * @param iterations the number of iterations to perform
     */
    public Wagging(ContinuousDistribution dist, Classifier weakL, int iterations)
    {
        setDistribution(dist);
        setIterations(iterations);
        setWeakLearner(weakL);
    }
    
    /**
     * Creates a new Wagging regressor
     * @param dist the distribution to select weights from
     * @param weakR the weak learner to use
     * @param iterations the number of iterations to perform
     */
    public Wagging(ContinuousDistribution dist, Regressor weakR, int iterations)
    {
        setDistribution(dist);
        setIterations(iterations);
        setWeakLearner(weakR);
    }
    
    /**
     * Copy constructor
     * @param clone the one to clone
     */
    protected Wagging(Wagging clone)
    {
        this.dist = clone.dist.clone();
        this.iterations = clone.iterations;
        if(clone.weakL != null)
            setWeakLearner(clone.weakL.clone());
        if(clone.weakR != null)
            setWeakLearner(clone.weakR.clone());
        if(clone.predicting != null)
            this.predicting = clone.predicting.clone();
        
        if(clone.hypotsL != null)
        {
            hypotsL = new Classifier[clone.hypotsL.length];